New book sheds light on human and machine intelligence

A single brain cell cannot think by itself, but when it’s connected with millions of other cells, that network is capable of everything from deciding what’s for dinner to contemplating the origins of the universe.

How this thinking “mind” comes about in both human and artificial intelligence systems is the subject of the new book “The Emergent Mind: How Intelligence Arises in People and Machines,” co-authored by Stanford cognitive psychologist Jay McClelland.

McClelland, the Lucie Stern Professor in the Social Sciences and professor of psychology in the School of Humanities and Sciences, is one of the pioneers of neural network models that gave rise to AI. For this book, he partnered with Gaurav Suri, who earned his doctorate in psychology from Stanford and is now an associate professor of psychology at San Francisco State University, to explain what we now know about how the mind works in humans and in machines.

We spoke with McClelland about how neural networks have changed the concept of the mind, how that knowledge impacted the development of AI, and, in turn, how AI influences what we know about ourselves.

This Q&A has been edited for clarity and length.

Why did you decide to write a book about how intelligence arises in humans and AI?

My co-author and I have the shared conviction that most people think about their mind as a special kind of entity that exists in a conceptual or experiential realm—and that maybe even comes from some sort of divine stuff. Research in the psychological and brain sciences, however, supports a different view—one we wanted to share with a wide audience.

The key idea is this: Whenever we have a perceptual experience, react to a situation, or feel a certain way about something we observe, that is all coming from processes taking place within the very rich system of interacting neurons in our brains. In the book, we describe models that we and others have developed that illustrate how this happens.

We called the book “The Emergent Mind” because we wanted to capture the idea that the mind does not exist as a unitary thing but arises from the interactions of very simple parts, none of which are themselves intelligent.

How have neural networks, like the models you’ve developed, deepened our understanding of how the mind works?

The models show how the experiences that we have—the thoughts, perceptions, and the decisions we make and actions we take—can be the product of the confluence of multiple factors, carried by signals from different neurons, that are operating behind the scenes. These are not accessible to our conscious mind, but they give rise not just to our behavior, but also to our experience itself.

Also, we’ve illustrated how the interactions among the neurons can be heavily shaped by our experience. Since we don’t perceive how our neural activity has been shaped, we tend to think that our experience arises from direct interaction with the real world as opposed to the effects of all these influences that arise from our past experiences and how they shape our reactions and perceptions.

Adopting a neural network framework for understanding the mind will ultimately lead us to be better able to think reflectively about what we think we know; how we judge others and our own views; and how we should live with each other and with our AI technologies because in some ways, they are very much like us.

What do you make of the recent advances in AI? Do you see them as approaching the level of human intelligence?

In the book, we explain how many of the successes of AI have been achieved because they rely on neural networks very much like the ones we use to model our human minds. But one of the important things that we talk about is how we’re not the same as our artificial systems. In our view, we are still smarter than our machines in many ways.

I take the view that by pursuing the effort to understand the human biological mind—as we’re doing in the field of neuroscience and interdisciplinary research on the brain basis of cognition—we will ultimately have a deeper understanding not only of ourselves, but also of how to build better machines.

So just as the original AI ideas have been supplanted by our better understanding of neural networks, we may learn more about how to build better AI systems and how to better understand ourselves.